perf: add basic benchmark to demonstrate overhead

[abraithwaite]

Not a scientific benchmark, but this sends 1MM events through a source and destination to demonstrate the low overhead of the framework.

No serialization is done in the benchmark, so achieving these results in real world applications is unlikely if not impossible.

The results are that the source + destination combo processes sent and received events at a rate of 1MM/478ms or about 2MM events/s.

goos: darwin
goarch: arm64
pkg: github.com/runreveal/kawa/test
BenchmarkMem-10     3    479789639 ns/op
BenchmarkMem-10     3    470265097 ns/op
BenchmarkMem-10     3    468312402 ns/op
BenchmarkMem-10     3    441723403 ns/op
BenchmarkMem-10     3    478304306 ns/op
BenchmarkMem-10     3    481978250 ns/op
BenchmarkMem-10     3    464515333 ns/op
BenchmarkMem-10     3    478364931 ns/op
BenchmarkMem-10     3    478826166 ns/op
BenchmarkMem-10     3    483086972 ns/op
PASS
ok      github.com/runreveal/kawa/test  28.571s

16:59:43 $ benchstat ./bench.txt
goos: darwin
goarch: arm64
pkg: github.com/runreveal/kawa/test
       │ ./bench.txt │
       │   sec/op    │
Mem-10   478.3m ± 3%